Generator

Assistant Test Question Generator

Nobody tests an assistant properly because nobody wants to write eighty questions. Paste your sitemap and this drafts a test set from your own page structure, with the awkward cases included rather than avoided.

Ask about CreobotSee all tools

YOU GIVE ITyour sitemapreadcomputereturnSITEMAP BECOMES A TEST SET????YOU GET BACKa test set drawn from your own pagesRuns in your browser. Nothing is sent anywhere.
  • Built from your pagesNot a generic list
  • Includes the edge casesWhere assistants fail
  • Re-run after changesA regression set

Convert it

Paste it here

Your file

These are drafts shaped from your URLs. Read them, fix the wording, and keep the set — its value is running the same questions again after every change.

How to use it

The edge cases are where assistants actually fail

Questions drawn from your pages mostly pass. Negation, compound questions and ambiguity are where behaviour falls apart, and they are exactly what nobody thinks to test.

Keep the set and re-run it

The value is not the first run, it is the second. A question that passed last month and fails now tells you a change broke something quietly.

Negation is the most reliable failure

Ask what a product does not do and most assistants answer what it does. It is a one-line test and it fails more often than anything else here.

Score citations by opening them

An answer marked correct with a citation nobody checked is not evidence. The link has to actually contain the claim.

Questions

They are drafts shaped from your URL structure. Read and fix the wording. The structure and the edge cases are the part worth keeping.

Roughly a minute per question including scoring. A forty-question set is under an hour and finds more than a week of informal poking.

The edge-case section, yes, on any public widget. The page-derived section needs their sitemap, which is public too.

Want this run against your whole site?

Tell us the page your assistant lives on and the questions you care about. A person runs them and sends back the raw answers.